Monitorare SQL Server può diventare un affare decisamente costoso. Certo, sul mercato ci sono moltissime soluzioni a pagamento, ma che fare se le istanze sono molte e i soldi sono pochi? In questa sessione combineremo diversi strumenti open source (InfluxDB, Telegraf , Grafana, DbaTools and many more) per raccogliere metriche di performance significative, analizzare i dati raccolti, creare degli alert per gli eventi critici, fare troubleshooting dei problemi e pianificare le risorse per il futuro. Raggiungimi in questa sessione e vedrai che il monitoring non è un business per milionari. By Gianluca Sartori

1. CLOUD
CONFERENCE
ITALIA
2019

2. SPONSOR

3. MONITORING SQL SERVER WITHOUT BREAKING THE BANK
Gianluca Sartori

4. Gianluca Sartori
Independent SQL Server consultant
SQL Server MVP, MCTS, MCITP, MCT
Works with SQL Server since version 7
DBA @ Scuderia Ferrari
Blog: spaghettidba.com
Twitter: @spaghettidba

5. AGENDA
• What we need
• What the market offers
• What SQL Server offers
• What the community offers
• The TICK stack
• How you do it

6. WHAT WE NEED

7. MONITORING
• Troubleshooting
What happened tonight at 03:40?
• Tune performance
Which queries consume most resources?
• Capacity Planning
How fast are my databases growing?
• Baselining
Is the system behaving normally?
• Alerting
Hey, look: something’s wrong here!

8. WHAT THE MARKET OFFERS
Buy a product
• Big value
• Big money $$

9. WHAT SQL SERVER OFFERS
 Use SQL Server built-in tools
• SQLTrace
• Extended Events
• Alerts
• Activity Monitor
• SSMS Reports
• Data Collector

10. SQL SERVER DATA COLLECTOR
 PROS:
• Included in SQL Server
• Offers some reports
• Good for troubleshooting
 CONS:
• Unreliable
• Slow
• No alerting
• Hard to customize/extend
• Deprecated

11. WHAT THE COMMUNITY OFFERS
 Generic monitoring tools
 Complete solutions
• Nagios
• Zabbix
• Cacti
 Tools
• Elasticsearch
• Kibana
 SQL Server monitoring tools
 Complete solutions
• SQLWatch
• TICK stack
 Tools
• dbatools
• ExtendedTSQLCollector
• XESmartTarget

12. THE TICK STACK

13. THE TICK STACK
Open source
Cross platform

14.  Time series database
 Written in GO
 Designed to store time series data
 Offers two query languages
• InfluxQL - similar to SQL
• Flux - inspired by Javascript / LINQ

15.  Data collection agent
 Can monitor several information sources:
• Operating system
• Databases
• Network
• Message Queues
• Apps
• …
 Works with plugins
 Over 100 plugins
• Apache
• Cassandra
• Docker
• Elasticsearch
• Graylog
• ….
• SQL Server!

16.  Administration interface for Influxdb
 User Interface for Kapacitor
 Can be used to explore data
 Can create dashboards to visualize data
 Can set up and manage alerts in Kapacitor

17.  Data processing engine
 Can publish data to topics
 Handlers subscribe to topics and perform actions
 Alerts can be sent to several channels, using plugins
• Email
• Slack
• …

18.  Monitoring / analytics platform
 Native Influxdb support
 More data sources supported via plugins
 Easy dashboard creation
 Multiple built-in / community dashboards

19. COLLECT STORE
PROCESS
QUERY
DASHBOARD
ARCHITECTURE

20. HOW YOU DO IT

21. 1. Download Influxdb
2. Unzip the archive (c:tick maybe?)
3. Run influxD.exe
• Open port in FW (default 8086)
• Run it as a service maybe?

22. Ever used srvany? You get why NSSM then…
1. Download NSSM
2. Unzip the archive (c:tick maybe?)
3. nssm install influxdb

23. 1. Download Telegraf
2. Unzip the archive (c:tick maybe?)
3. Generate and edit the config file
• Log
• Input
• Output

24. 1. Download Chronograf
2. Unzip the archive
3. Install as a service
4. Open in browser
• Create a connection
• Create dashboards
• Explore the data

25. Influx DB entity SQL DB entity
Measurements Tables
Tags Columns (indexed)
Fields Columns (not indexed)
Points Rows
•Tags are indexed, and are used to filter and aggregate data
=> view CPU usage (the Field) by host (the context)
•Fields are not indexed and should only be aggregated.
A field should represent a metric.
=> CPU usage

26. 1. Download Grafana
2. Unzip the archive
3. Install as a service
4. Open in browser
• Create a connection to Influxdb
• Create (or downlad) dashboards
• Explore the data

27. 1. Configure alerts by dashboard
2. Multiple channels available
• Email
• Slack
• Teams
• Telegram
• …

28.  Continuous Queries
• Scheduled queries in InfluxQL
• Provide downsampling and aggregation of data
• Can write to a different retention policy
CREATE CONTINUOUS QUERY "cq_basic" ON "transportation"
BEGIN
SELECT mean("passengers") INTO "average_passengers"
FROM "bus_data" GROUP BY time(1h)
END

29. TICK STACK: RECAP
 Troubleshooting
Grafana dashboards, standard sqlserver dashboard + custom dashboards
 Tune performance
Grafana dashboards, standard sqlserver dashboard + custom dashboards
 Capacity Planning
Influxdb continuous queries, custom Grafana dashboards
 Baselining
Influxdb continuous queries, custom Grafana dashboards
 Alerting
Kapacitor or Grafana

30. RESOURCES
• Influxdb, Telegraf, Chronograf, Kapacitor
https://www.influxdata.com/
• NSSM - Non Sucking Service Manager
https://nssm.cc
• Grafana
https://grafana.com

31. GRAZIE!